Thèse Adaptation des Attendus Génomiques aux Population d'Espèces Partiellement Clonales Modélisation Inférences et Tests Empiriques H/F

Doctorat.Gouv.Fr

Établissement : Université de Montpellier
École doctorale : GAIA - Biodiversité, Agriculture, Alimentation, Environnement, Terre, Eau
Laboratoire de recherche : MARBEC - Biodiversité Marine, Exploitation et Conservation
Direction de la thèse : Sophie ARNAUD-HAOND ORCID 0000000158148452
Début de la thèse : 2026-10-01
Date limite de candidature : 2026-05-07T23:59:59

La clonalité partielle (i.e. combinaison de sexualité et de reproduction clonale) est fréquente dans le vivant et concerne des espèces à enjeux majeurs (pathogènes, invasives, plantes, espèces structurantes terrestres et marines). Ce mode de reproduction influence fortement l'écologie et l'évolution, et doit être intégré aux analyses de génétique des populations. Les modèles théoriques développés dans les ANR Clonix/Clonix2D ont affiné les attendus génotypiques et génétiques sous divers taux de clonalité. Ces avancées doivent être consolidées sous différents scénarios démographiques et transformées en méthodes d'inférence accessibles. Parallèlement, l'essor des génomes complets d'organismes non modèles partiellement clonaux pose des défis, car les pipelines bio-informatiques standard biaisent les signatures de clonalité (équilibre d'Hardy-Weinberg, déséquilibres de liaison). La thèse visera (1) à densifier les bases théoriques, (2) développer des modèles d'inférence par apprentissage supervisé, (3) concevoir une méthode adaptée aux génomes complets. La preuve des concepts développés se fera sur une phanérogame marine et deux coraux Scléractiniaires, espèces ingénieures d'écosystèmes marins clés côtiers et profonds.

Partial clonality, the coexistence of sexual and clonal reproduction within the same species, is widespread in the tree of life [1, 2]. Now, a large fraction of eukaryotes are partially clonal, including many taxa of high societal, environmental, and economic importance (human and crop pathogens, invasive species, most photosynthetic organisms, and numerous marine and terrestrial ecosystem engineers). Partial clonality deeply changes species demography, evolution, life-cycle, and interactions with their environment [3-5], with direct implications for population genetics, when theoretical framework and methods are developed for pure sexual or asexual organisms [6, 7]. First, genotypic diversity long used as a proxy for the rate of clonality, has proved unreliable due to sampling sensitivity [6]. Second, simulations predict deviations from Hardy-Weinberg equilibrium (HWE) in partially clonal populations where equilibrium is reached extremely slowly across evolutionary time-scales [8, 9]. Clonality also affects the very concept of effective population size (Ne): higher clonality increases the number of identical individuals, with direct consequences for genetic diversity and adaptive potential[10]. Recent modelling [7, 11] highlights the general need to refine theoretical expectations and inference methods, notably by feeding learning approaches with simulation data across a range of demographic and mutational scenarios.

At the genomic scale, challenges are even greater. Large datasets inevitably include genotyping errors, imposing curation through filtering steps, but existing bioinformatic pipelines-developed under assumptions of panmictic sexual populations-systematically lead to bias in data analysis, and their interpretation under partial clonality. For example, most SNP callers and filters rely on HWE expectations or minimize linkage disequilibria, erasing all or part of the clonal signature proposed in the literature [7, 9, 12] while producing high-quality datasets. The very signals needed to infer clonality and related demographic parameters are thus lost with the tools available thus far. To address this, theoretical expectations must be consolidated by integrating knowledge from simulations and analytical models (Markov chains, Stochastic Differential Equations, ...) to formalize links between clonal rate, effective population size, and genetic diversity. These theoretical results are required to fuel supervised machine-learning methods and develop a robust inference framework.

This project builds directly on advances from the ANR Clonix and Clonix2D projects [7, 13], which developed foundational theoretical frameworks and initiated supervised inference approaches. It will extend these models to genomic data using available datasets (e.g. 200 coral genomes, 150 Zostera marina genomes from Biocean5D). Together, these provide a unique opportunity to consolidate theoretical models, adapt bioinformatic pipelines, and bring inference of clonality into the genomic era with a significant contribution to the understanding of the ecological dynamics of population of species of special ecological interest (threatened engineer species such as seagrass and corals; non-indigenous species; pathogens etc...).

The objectives of the thesis will be i) to understand how to aggregate genomic and whole-genome data for partially clonal species and to use this knowledge to better preserve the signatures required for methods inferring population clonality rates and ii) to contribute to understanding the effects of clonality on the population genetics of partially clonal species and on their genomic specificities, in order to propose avenues for adapting existing genomic methods developed for strictly sexual organisms.
To achieve this, the thesis will 1) enrich reference datasets of numerical expectations under increasing clonality through simulations and stochastic models across different demographic scenarios ; 2) develop inference methods based on supervised machine learning, tailored to the specificities of genomic and whole-genome data and formally integrating expected linkage disequilibria in such populations ; 3) improve high-throughput genomic data preprocessing and filtering tools, and apply them to high-density SNP datasets (e.g., from RADSeq or whole-genome sequencing) already available within the host team.

1.Improving theoretical expectations in partially clonal population genetics
The first step will extend models from Arnaud-Haond et al. [13] and Stoeckel et al. [7] to refine inference of clonality rates and clarify links between effective size, demographic changes, and parameters describing genotypic/genetic diversity. Simulations and formal models (e.g. SDEs) will generate datasets with known parameters to test inference methods through cross-validation to define methodological guidelines, and apply these approaches to datasets already available (part 3).
2. Supervised machine learning
Inference models based on supervised learning (ABC, Naive Bayes, Random Forest, genetic algorithms) will be coupled to numerical model outputs and reference datasets, first validated on simulated data and then applied to empirical data. Building on preliminary work (Rouger, unpublished), the goal is to jointly infer clonality rate, mutation rate, effective size and current population dynamic in agreement with theoretical expectations. This work will be supported by collaborations with CRBE (Toulouse).
3.Filtering empirical whole-genome data
The third step will address errors and biases in current genome assembly and SNP filtering pipelines. The objective is to correct sequencing errors without erasing clonality signatures, notably linkage disequilibria. The candidate will analyze biases in existing tools (e.g. GATK, NGS) and propose methodological improvements. Throughout these steps, methods will be applied to existing datasets from three marine engineer taxa: seagrasses (Zostera marina) and deep-sea corals (Lophelia pertusa, Madrepora oculata). These species play key roles in marine ecosystems, are highly sensitive to climate change, and as for all other species, their mating system is expected to strongly influences resilience and colonization dynamics. Understanding their genetic and genomic responses is essential for biodiversity preservation.